Digital Interconnects

[The Devil]

Audiophoolery is what make the hi-fi business go around. People fooling around with valve CDPs, poncy graphic equalisers, and the like. The truth sounds "wrong", so let's try and muck about with it. There's one born every minute.

I think cable/connectors are important enough to try to stop them vibrating.

 

[MartinC]

Signal travels at about 200 000 000 m/s, clock is about 3 000 000 Hz which implies a wavelength of about 66m. On the face of it theories regarding cable length and reflections don't seem likely.

I suppose there will be much higher frequency content to generate the sharp square wave transitions on the digital signal...

 

[Paul Ranson]

I said 'clock' not the max frequency of the spectrum of the signal. The question is whether reflections due to bad termination at the transmitter can cause confusion at the receiver and what the cable length to maximise this putative effect would be.

I think more theory from 3DS would be useful before speculating further.

Paul

 

[3DSonics]

Hi,

I think more theory from 3DS would be useful before speculating further.

I'll try, I'll have to pull info in most of which not at hand to answer comprehensively.

First, the way S/P-DIF (AES/EBU) work is by serieally encoding the data (total of 64Bits per sample period) using a F/2F (Bimark Phase encoding). This means the fundamental clock is at 2 X 64 X fs or in case of CD at around 5.6MHz.

The clock in the DAC is recovered by a PLL which syncronises to the source datastream. In order to do so with good reliability the system use the "edge transtition" when the signal exceeds a certain value (in case of the cirrus logic receivers commonly found at 200mV).

The Rise time (in other words the "width" of the edge) in S/P-DIF is commonly somewhere between 5nS to well past 20nS. If for any particular reason the point where the edge transtitions the 200mV point is not the same, relative to the "nominal" point where it should be, we have a variation in timing, which will be attenuated by the PLL only above around 10KHz modulation frequency.

In order to avoid degrading a 16-Bit/44.1KHz sample rate signal at conversion (measureably) we require a jitter level of less than 150pS, audibility levels have IIRC (need to check references) been established in the 500pS+ Jitter range.

From all the above and the knowledge that the connectors (RCA & XLR) are not constant impedance we realise that some of the signal send down the cable will (MUST) be reflected and will arrive sufficiently delayed that it will shift the edge transtition. The same will happen of course with cable that has a relative high impedance screen (simply due to noise currents circulating in the screen).

So, if we shift the switching point in a non-constant fashion by a rather miniscule amount we can generat measurable degradation.

BTW, in 1nS Light travels 12" (appx.) and most cables have less than 80%c propagation speed. In 100pS Light travles only about 1.2"....

That so far for the introduction.

Ciao T

 

[oedipus]

Hi,

I'll try, I'll have to pull info in most of which not at hand to answer comprehensively.

<blah, blah, blah>

That so far for the introduction.

Ciao T

Well, before you get too carried away with all this Electrical Engineering....

You might want to consider this: 75Ohm cable is 75Ohm cable whether it has a RadioShack or Nordost label on it...

Let me put this another way: the direction your heading in seems to demostrate that any 75Ohm cable is adequate...

You agree with this?

If so, you can move on to connectors

 

[3DSonics]

Hi,

You might want to consider this: 75Ohm cable is 75Ohm cable whether it has a RadioShack or Nordost label on it...

Well, there may any number of material differences. First, the propagation speed varies depending on the dielectric. Secondly the precision with which the 75 Ohm is achieved and held along the cable may vary. Thirdly, the resistance and Impedance of screen (ground) may vary, giving different levels of (jitter inducing) ground fault current derived voltages.

However, you may have noted that I recommend short pieces of CT/FT-100 which costs at Maplin £1/m, I fail to see the link between cost and electrical performance of cables which cause a difference in sound.

Muddeling up the issue by bringing up "expensive cables" is irelevant in the context and indicative of either poor logical thinking or deliberate intent to mix unrelated issues.

Let me put this another way: the direction your heading in seems to demostrate that any 75Ohm cable is adequate...

You agree with this?

No, not as such blanket statement. Depending on cable length, context of application etc "just any 75 Ohm CI Cable" is insufficient. Luckily enough even very high performance 75 Ohm CI Cable is available rather inexpensively.

If so, you can move on to connectors

We have been there.

You seem to want to demnstrate that "expensive cables" make no difference. For that to be true it is essential of course that NO cable makes a difference and that view is wrong, patently so.

I suggest you seperate the issues of the cost cables in the high end context (which tends to be high due to hand assembly in small numbers plus a veriety of markup at distributors and dealers and a mentality of "let's charge what the market will pay" - in short - capitalism) and the actual electrical and possibly esotheric reason that create reasons finally resulting in subjectively percieved differences.

Ciao T

 

[wadia-miester]

I have at least 6 cables which are all true 75 Ohm units, they do all have a different sonic sigature and yes they all have the same connectors on them.
3 of them are constructed of indentical material, 3 are not.
Though that opens a whole new can of worms LOL!!!!

 

[3DSonics]

Folks,

Digital Cables part two....

We left below at:

BTW, in 1nS Light travels 12" (appx.) and most cables have less than 80%c propagation speed. In 100pS Light travles only about 1.2"....

Now, let us assume a S/P-DIF standard Digital cable, 1m long, 50% (of c) propagation speed (due to PVC insulation) and with RCA Plugs.

Sadly, no RCA sockets (except possibly WBT nextgen, but I'm not taking their word) have anywhere near 75 Ohm CI, many will be closer to 30 Ohm CI.

So a proportion of somewhere between 20 & 50% of the signal reaching the far end of the cable will be reflected. The round trip of this reflection back to the source (where another non-75R-CI connector reflects a good deal of the reflection) will take around 6 - 7nS so the now re-arriving reflection which still has an appreciable proportion of the signal strenth of the desired signal arrives well within the leading (and triggering) edge of the signal.

And let us remember, our signal will "trigger" the recevier (if it is a cirrus logic one) when a +/-250mV square wave exceeds 200mV.

The chances that the interference will effect the trigger window is not small, as with (say) 10nS rise time the triggering will be just around where the reflection arrives.

NOW, if we had an unchanging signal with a stable clock, that would matter FA, as the distrubances would be present on EACH AND EVERY transttion. But in S/P-DIF we change the clock to signifiy bits, so the disturbances on the triggering edge are signal dependent.

Now, if we make our cable sufficiently long that the "roundtrip" of relections take much longer than the triggering on the edge, but not so long as to interfere with the next triggering edge we will have eliminated this source of jitter (jtter due to interchassis differences, RFI etc remain).

Equally, if keep the cable very short (as I said, < 10..12" and ideally with a very high propagation velocity) we will not materially effect the edge triggering, as the reflection arrives almost instantanious back, a further advantage of very short cables is that other jitter soures (interchassis noise currents generating fault voltages in the non-zero impedance ground return, EMI, RFI et al) are also kept to an absolute minimum.

Anyway, that is more or less abstracted and simplified the reasoning behind "make digital S/P-DIF cables either VERY SHORT or comparably VERY LONG.

Of course, some listeners may prefer a system with more jitter and of a certain spectrum, which normal length digital cables will readily provide, but in a manner and fashion that remains extremely unpredictable from system to system and cable to cable.

The best solution is to avoid S/P-DIF entierly and the secondbest is to switch to AES/EBU connections as the XLR connectors (without metal shell) is pretty close to 100 Ohm CI and as the 5V Peak-Peak Level and balanced input to the receiver confine the triggering to a much narrower part of the leading edge of the signal. This still does not ensure that there are no other interactions of course, but it raises the bar aginst such things considerably (as do +20db @ full scale pro-style balanced connections do compared with 2V @ full scale consumer single ended connections).

Ciao T

 

[Paul Ranson]

Surely if an edge is interfering with itself then the effect must be the same for every edge?

And it's a big jump from jitter on an edge to jitter in the derived sample clock amd on to implied audibility. IMO.

And all the implied problems go away with the use of BNC connectors, pure S/PDIF compatibility being retained with a cheap BNC/Phono adaptor...

Paul

 

[3DSonics]

Hi,

Surely if an edge is interfering with itself then the effect must be the same for every edge?

Depends. If the stream is just 1/0/1/0/1/0.... it will be, if it varies depending on data (11/00/11/0/1/0/1....) the interference will not neccesarily the same.

And it's a big jump from jitter on an edge to jitter in the derived sample clock amd on to implied audibility. IMO.

Well, the chain goes like this:

1) Reflections in cables can cause jitter after the PLL easily in the region of significant fractions of 1nS, plus as said, other mechnasims past refections exist that result in the same.

2) Jitter of significant fractions of 1nS is normally considered at least POTENTIALLY audible.

Thus, reflectiosn in typhical S/P-DIF cables of typhical length may result in sufficuently large degrees of jitter to be potentially audible. Do you dispute this chain of reasoning?

And all the implied problems go away with the use of BNC connectors, pure S/PDIF compatibility being retained with a cheap BNC/Phono adaptor...

I agree partially, using BNC connectors AND ensuring correct termination in source and receiver will remove ONE of the problems (reflections), but leave others (limited signal level and poor rise time and poor resilience to groundfault current induced problems) untouched.

Going for fully transformer coupled balanced inputs and outputs using the AES/EBU standard is much more likely to address the whole lot (in analogue BTW too).

One of main reasons for the seemingly disproportionate impact of cabeling that SHOULD NOT impact on sound in HiFi gear is the ill conceived single ended interface (it's the main reason why mains cable actually make a difference!!!).

Ciao T

 

[Paul Ranson]

Do you dispute this chain of reasoning?

Well, yes, a bit(!)...

You haven't shown that reflections in cables, or noise, or inherent jitter in the source, cause post-PLL jitter, you've merely claimed it. It seems to me to be a relatively straightforward engineering problem to avoid this.

You have derived your 1nS jitter post-PLL, which is in the S/PDIF clock, and assumed the same jitter will be passed on to the sample clock. Apart from the issue of good DAC engineering this doesn't seem a generally obvious assumption.

It would seem relatively straightforward to take a DAC and watch the jitter artifacts on the analogue outputs change with digital interconnect, and easy to provoke the extremes with deliberate mismatches to validate the experiment.

Paul

 

[3DSonics]

Hi,

Well, yes, a bit(!)...

Hmmm, do you dispute that it can happen and due to the design does happen in most "hifi" and "high end" gear, regardless of price or do you just think "competent design" should deal with it?

You haven't shown that reflections in cables, or noise, or inherent jitter in the source, cause post-PLL jitter, you've merely claimed it. It seems to me to be a relatively straightforward engineering problem to avoid this.

I agree that is is fairly trivial to avoid, HOWEVER it is NON-TRIVIAL when manufacturing and designing mains powered equipment that conforms with electrical safety standards, the usual consumer interfaces and so on. It is still doable, but for whatever reasons rarely done. Funnily enough, the mid 1980's Marantz DA-12 DAC used a discrete low noise secondary PLL for clock recovery and opto insulation between digital section and DAC, something rarely seen in this modern age.

You have derived your 1nS jitter post-PLL, which is in the S/PDIF clock, and assumed the same jitter will be passed on to the sample clock.

No assumption. Fact for the majority of commercial DAC's simply because of the chipsets and circuits used.

Very few (and then only extremely expensive High End DAC's) have a slow secndary PLL (examples are db Technoclogy [or whatever their name now is], Audio Synthesis) or a memory buffer (Chord) to allow them to filter the clock further or just substitute it with a fresh one.

More DAC's use ASRC's but these strictly speaking do not remove clock jitter, they redistribute it spectrally and encode it unremovable in the signal.

It would seem relatively straightforward to take a DAC and watch the jitter artifacts on the analogue outputs change with digital interconnect, and easy to provoke the extremes with deliberate mismatches to validate the experiment.

HFN IIRC did publish a few Digital Interconnect reviews years back with measurements included. They did not repeat this very often, perhaps because some really expensive and fancy cables measured very poorly.

NOW, to put you on the spot, do you have any problems with my recommendation for a good sounding RCA based digital interconnect?

That is keep it very short and use CT/FT-100 75 Ohm Satellite coax from Mapli at £ 1/m and inexpensive low metal content RCA Plugs.

And how about my reasoning as to WHY this makes a good cable?

I suggest that minimal ground impedance due to minimal length and large crossectional area as well as low inductance for the screen/ground minimises groundfault voltages, minimal length also eliminates or minimises problems with delayed reflections (be they real or not).

How about it? What is wrong with my digital interconnect suggestion which costs probably around 2 damp squid in materials and 10 Minutes max soldering and applying heatshrink?

Ciao T

 

[leo]

I use 1.5m of Belden 1505A terminated with 75ohm BNC's
It does the job ok

 

[oedipus]

NOW, to put you on the spot, do you have any problems with my recommendation for a good sounding RCA based digital interconnect?

Yes. The problem is with your claim that it sounds "good", rather than "just like every other 75Ohm cable".

 

[The Devil]

Who cares about the cable, when a Cello Palette lurks downstream, waiting to mangle the music?

 

[3DSonics]

Hi,

Yes. The problem is with your claim that it sounds "good", rather than "just like every other 75Ohm cable".

So you mean that even though it may cause high enough levels jitter to be audible AND even though groundfault currents may push such jitter levels into the 10nS+ region, when compared to the cheap & short cable they still MUST sound sound the same, because you say so?

Have your dogmatic "anti high end cable" escapades made you really so deaf (and blind) to the very real issues involved?

Ah well, suits you sir.

L8er T

 

[Paul Ranson]

NOW, to put you on the spot, do you have any problems with my recommendation for a good sounding RCA based digital interconnect?

No, none at all. I'm just not convinced about audible jitter caused by termination issues. And while it's easy to say 'ground fault currents' and easy to see why they would cause jitter, as would any noise, I'm also not convinced that they exist in real equipment. I'm pretty sure that even my old Arcam Alpha has a transformer coupled S/PDIF output, my M-Audio Audiophile 24/96 definitely does. I haven't looked inside my relatively cheap DVD player.

Paul

 

[oedipus]

So you mean that even though it may cause high enough levels jitter to be audible AND even though groundfault currents may push such jitter levels into the 10nS+ region,

Just demonstrate with EE (and give us the math as proof), or provide a set of measurements to show that the claims you are making are true and do not need to be qualified by may. If you need to use words like "may" you are simply guessing...

Have your dogmatic "anti high end cable" escapades made you really so deaf (and blind) to the very real issues involved?

You can keep presenting anecdotes on digital signals and various tidbits from EE until the cows come home, what you have to do is show that "the very real issues" you are refering to cause an audible distortion in the signal coming out of the DAC.

I could literally be deaf and blind. Nevertheless, I would still be able to tell that as things stand, you are giving lots of disjoint facts, but they don't fit together to make a case.

 

[3DSonics]

Hi,

No, none at all. I'm just not convinced about audible jitter caused by termination issues.

I accept that, however, can you reliably, beyond any reasonable doubt exclude termination issues from the picture? (If so, proof please).

IF you cannot reliably exclude them, would you then agree that these strictly hypothetical issues (1) would be SIGNIFICANTLY reduced by selecting a cable with one of the higest propagation velocities readily available and making it as short as possible?

(1 hypothetical in the context of this discussion anyway, I have reason to believe they are not hypothetical, but I do not wish to write and defend a doctoral thesis on the subject)

And while it's easy to say 'ground fault currents' and easy to see why they would cause jitter, as would any noise,

Okay, so IF noise is present we have jitter? Yes?

I'm also not convinced that they exist in real equipment.

A 'scope, 100R resistor and an earth from a local outlett may rapidly disabuse you of your conviction.

I'm pretty sure that even my old Arcam Alpha has a transformer coupled S/PDIF output,

Of course it does. BUT (and here comes the double whammy), the transformer secondary is NOT left floating but connected to the chassis (you have to, for compliance with electrical safety, FTC and anti EMI regulations and I agree, it's a daft thing to do).

So, if there is a material ground fault current in the chassis (2) then this is likely to flow in the cables ground conductor one way or the other.

(2 on mains powered equipement ground fault currents and other noise caputred by the grid and cabling acting as aerials are practically completely unavoidable and usually in the region of several 100uA in the BEST cases and near legal limits [10mA IIRC] in the worst)

Now, I ask again, based on what has been submitted, can you reliably exclude mistermination caused reflections AND the presence of ground fault currents sufficient to cause problems IN S/P-DIF links? Please substantiate by reference or empirical study.

Ciao T

 

[zanash]

I think you'll find that the transformer coupled output of the spdif is more to do with reflection elimination rather than jitter. Rhere are some intresting articles out on the web to do with connection type and signal reflection....oh yes thes are not hifi but tellecomunication aplications.